*2.8. Gene Ontology (Enrichment) Analyses*

The clusters based on the 2500 most variable genes were further analysed to visualize overrepresented biological terms using the whole GO and KEGG term space from Section 2.4 as a background. We explored enriched pathways within the expression clusters using ClueGo v2.5.0 tool [95], a visualization plug-in implemented in the Cytoscape v3.6.0 environment [96]. Term overrepresentation was estimated by an enrichment test based on the hypergeometric distribution followed by Benjamini–Hochberg adjustment for multiple testing. An annotation network was built with the ClueGo plug-in from kappa scores, which reflect the associations between genes and GO and KEGG terms. Network specificity was set between 3 and 12 GO hierarchy levels, and term selection was set to a minimum of 3% genes per cluster. Kappa score threshold was set to 0.3, and we allowed GO parent-child term fusion. Moreover, we explored the network with the MCODE algorithm [97], implemented as a Cytoscape plug-in, to detect densely connected regions or hubs in the network. Those hubs were found in the network establishing a degree cut-off of 2 for network scoring criteria, without including loops. Option Fluff was selected and parameters for Cluster Finding panel were set at 0.1 and 0.2 for node density and node score cut-off, respectively, a minimum of 2 edges per node of cluster cores (K-Core) and a maximum depth of 100.
